Steve Dine: Modern Cloud Architecture & Mistakes To Avoid When Moving To the Cloud
In this episode, Wayne Eckerson asks Steve Dine about the approach needed to migrate to the Cloud and architecture required to run analytics in the Cloud. Steve Dine talks extensively about the pitfalls to avoid during cloud migration and finishes off by saying that even though security is a big issue, most organizations will have part of their analytics architecture on the cloud in the time-span of next two-three years. Steve Dine is a BI and enterprise data consultant and industry thought leader who has extensive experience in designing, delivering and managing highly scalable and maintainable modern data architecture solutions.
Steve combines a strong business acumen with hands-on technical experience and is committed to providing value by helping companies transform their data assets into a competitive advantage. His experience spans Healthcare, Manufacturing, Financial Services, Life Sciences, and Retail. He is currently President at Datasource Consulting.
- Cloud might soon become a commodity like electricity
- Companies jump into cloud migration only to learn later that there are many compliances in their industry or even in a country that doesn't allow storing data in a cloud
- Find out what are users trying to do with data- are they sampling the data or are they downloading on their desktop for reporting? It affects the Cloud architecture decision
- Testing data transfer speeds at different times a day while moving your data to the Cloud helps in calculating how long could the transfer take
- Hybrid cloud can have two meanings: A). A cloud architecture which is internal to the organization but managed externally. B) The other is a hybrid cloud architecture in which one would have some of their analytic architecture on-premise and some of it on the cloud
- If you have a large contingent of data users and lots of compliance issues, it's better to have a hybrid architecture
- If an enterprise is spread all over the globe, it is better to store the data locally than on the Cloud as it will have low-latency and can be queried easily
- When data gravity is on-premise, BI can go to the Cloud before the data does
- In the Cloud, you tend to leverage object-based storage as a lot of organizations don't have object-based storage. It's cost-effective to store data
- You can spin up more resources on the Cloud as it can be architected for distributed and parallel environments
- Not having a strong plan on governance can hurt your Cloud migration plans
- The first mistake that people do when moving to the Cloud is not understanding the economics of the Cloud
- Lift and shift to the Cloud may or may not work in all cases. It may be cost-effective but not the best option for architecture
- You need to be worried about vendor-lock-in the Cloud. The more services you use in the Cloud, the more you get locked in
- To avoid lock-in, you should architect in the beginning to understand what to move and how to move
- Customer analytics needs can be met faster in the Cloud when compared to on-premise
Below is one question and answer from the podcast
Wayne Eckerson: Going to the Cloud is a no-brainer, do you think this is really true?
Steven Dine: Well, I wouldn’t say necessarily that’s a no-brainer. For a lot of companies, there are many things to consider. I do think over the long-term Cloud will be prevalent how we architect it in our systems. There is an analogy out there about cloud being a public utility. Nicholas Carr wrote a famous article called "IT doesn’t matter" and then wrote a follow-up book about Cloud comparing it to electricity and how we there are companies who generate their own electricity. Now, electricity is a commodity and rarely do we have our own electric generation per organization. Compare the Cloud to that as a model in today’s world where most companies have data centers; at least larger organizations I’ll say. So in the future, will organizations have their own data centers or will they leverage cloud like a commodity?
So when we think about the cloud for analytics and data management, there are a lot of things that need to be considered. Sometimes the most important is compliance. Organizations a lot of times start putting their plans together about migrating to the cloud only to learn that they are in an industry or they have compliance regulations or even in a country which doesn’t allow storing data in a public Cloud. As I mentioned, there is compliance in healthcare, banking, government and a lot of different industries have requirements which public Clouds are not able to meet.
You definitely need to understand from a data perspective what you can do or what you cannot do in moving your data out of your local data center. You also have to consider the volume of data in your organization and where that data is located which is referred to as data gravity. Assessing organization readiness when migrating to the Cloud is one of the tough things to look at because moving larger volumes of data can be problematic. Even though networks have become faster and compressions have become better, still you can only move so much data over the network.
So there are two things to consider. a) the volume of data to be moved to the Cloud b)what are the usage patterns- are people using that data for analytics or they downloading the data and doing the analytics on their desktop tools
Depending on the type of services you are looking from the Cloud, the responsibility model changes. Many organizations justify the Cloud move through lowered costs.